A Problem For Direct Reference
According to the theory of direct reference, the meaning of a name (and possibly an indexical expression) is the object that is its reference—typically an ordinary concrete object. The weak version of the theory says that this object is at least part of the meaning of the name; the strong version says that the object is the whole of the name’s meaning. In either case we can say, according to a standard formulation, that the object enters the proposition expressed by the sentence containing the name: it is a constituent of the proposition so expressed. Just as senses or intensions or concepts are said to enter into propositions, so objects themselves can be said so to enter. Thus a singular proposition is said to be an ordered pair of an object and a property, not just an assembly of properties. Meanings can be partly composed of objects—extensions not intensions.
But objects have properties: some people hold that they are just aggregates of properties (bundles, sets). So aren’t properties a part of the package? If objects are contained in the proposition, don’t properties come along for the ride? If properties can be constituents of propositions in their own right, they can also feature in propositions as borne by the objects that instantiate them. Objects can hardly leave their properties behind when they step into propositions—they would not be objects at all if they did that. No, the object brings its properties into the proposition too—all of them (it doesn’t shed some and retain others). But if the object is the meaning, then these properties are also contained in the meaning. The parts of the object clearly accompany it into the heart of the proposition, and so do its actual properties; an object is a complex thing and its complexity survives in the warm environment of the proposition. The idea of direct reference is that the proposition contains at least the reference, but then it also contains what the reference contains—properties as well as parts. If it contains properties as detached from objects, it need not contain more than the properties in question (and their logical entailments); but if it contains objects, it will also contain all the properties that the object possesses. Thus the meaning of a name, so far from being simple and austere, is replete with properties—many more than under a classical description theory. The object itself guarantees that.
This has a rather startling consequence: all true sentences containing names are analytically true. For every property we can truly predicate of the object is already contained in the meaning of the name via the object named. The properties get uploaded into the proposition via the object and hence anticipate whatever might be truly predicated of the object. If only a single property occurs in the subject position, then we don’t have this consequence—there will only be one analytic truth generated by the name—but if the object brings all its properties along with it, then every true sentence containing the name will be analytic. That means every such sentence will be a necessary truth. Not an a priori truth to be sure, because speakers can’t be expected to know every property of the denoted object, but a truth that follows from meaning alone. This Leibnizian conclusion falls directly out of the direct reference theory, unless some way can be found to block it. It’s like putting a bunch of objects into a drawer: the objects will be in there but so also will their properties—the objects don’t stop having their properties just by being put in a drawer. Take a peek inside and you will see the properties lurking in there. The proposition doesn’t contain bare particulars but propertied particulars. If objects are literally bundles of qualities, then you will be putting such bundles into the drawer—or the proposition. This is not a consequence of the direct reference theory that I have ever seen acknowledged, but it appears inescapable.
But there is an even more startling consequence in the immediate offing, which I hesitate even to mention. Some may see in the derivation just rehearsed a welcome refutation of the direct reference theory, but what if I told you that the same result, or something very like it, applies no matter what the theory is? Glee might then turn to panic—or hysterical incredulity. Here is one way to put the point: when properties enter into propositions (i.e. sentence meanings) they too have properties—for instance, the property of being instantiated, or the property of being co-instantiated with another property in a certain object. For example, the property of being red has the property of being instantiated by British post boxes and also the property of being co-instantiated with roundness in a cricket balls. Properties thus bring with them other properties that are not analytic entailments of the given property; and this gives rise to what may be called surprise analytic truths—as that redness has instances and is the color of cricket balls. If the property itself has certain properties, then this property will bring these other properties along with it, thereby generating surprise analytic truths. Or consider senses: these too have assorted properties—having a certain reference, being currently grasped by the Queen of England, being sometimes too complex for humans to get their minds around, etc. So when a Fregean Thought contains such senses must it not also contain the properties of those senses? They come with the territory, just as in the case of objects and theirproperties. Why do we assume that only some properties of senses properly belong to the meaning of a sentence expressing them—why not all? By what magic might some properties of senses be incorporated and some not? Frege compared senses to aspects of references, but aspects are themselves rich with attributes—a given aspect might be an aspect of Mount Everest, or frequently found with another aspect, or apprehended by some individuals but not by others. So aspects of the aspect are going to find themselves wherever the aspect finds itself—lodged inside a sense. We are beginning to see the outlines of a general problem—the problem that whatever we choose to constitute meaning carries within it too much to be a meaning, as meaning is normally understood. We want meaning to be a thin slice of reality, so to speak, but the available theories always make it a good deal thicker than we bargained for. We could call this the thickness problem—and it threatens to undermine the notion of meaning as we have it. How do we stop too much of the world from getting into the proposition? How do we stop meanings from absorbing too much of reality? Any entities they comprise are bound to be carrying a cargo of extra baggage—to be overstuffed, overly inclusive. The entities will have features that make them (allegedly) suitable to function as meanings, but they will also have other features that are irrelevant to meaning: the features arrive in a package that can’t be broken apart. In the case of direct reference theory, we want the object to be in the proposition for various theoretical reasons, but the object presents other features that disqualify it from doing the job of meaning, notably the totality of its properties. We end up having to say that every true sentence containing a name is analytically true, and that the meaning of sentences containing names can never be fully grasped. It looks like we are headed for the conclusion that meanings cannot be made up of entities at all. But what else could they be? What else could anything be? Objects, senses, properties, functions, and images—all these are entities that bear more properties than meaning intuitively bears. So how could any of them be meaning? But then how could meanings exist at all? We are headed straight for skeptical paradox. Nor will use get us out of trouble, since uses too contain more than meaning does—vocal noises, laryngeal movements, events in the motor cortex, etc. Anything that exists in the empirical world has an abundance of properties that go beyond anything that could constitute meaning, so how could meaning be identical to anything like that? Empirical reality is just too thick to constitute meaning, too multilayered. And abstract entities such as we find in mathematics won’t work either: meaning isn’t mathematical, and anyway the problem will recur in a new guise—for mathematical entities are too ontologically rich as well. Propositions are sparse etiolated things, sliced very thin, but any facts we can mention, empirical or abstract, have too much depth, too much substance, too much thickness. The point is very vivid in the case of the direct reference theory, but apparently much the same problem afflicts theories that deal in more rarified entities. It looks like there is no way out, as a matter of principle. It looks like we have a problem, Houston. Meaning is having difficulty getting back down to earth. 
 References don’t present the same problem: they can be as rich and extensive as you like without generating unwanted analytic truths. This is because they are not construed as part of meaning in the Fregean picture, so there is no risk of them stuffing meaning with too much material. But senses are meant to be constitutive of meaning (which is commensurate with linguistic understanding) and so cannot accept any old properties into their constitution–yet they have many properties that go beyond what meaning intuitively contains.
 I am as shocked as the next man by the argument of this paper and I keep thinking there must be a simple answer to it, but so far I have been unable to find one. I put it out there with some misgivings. Perhaps I should announce that I thought of it while trying to interpret an obscure text by an unknown philosopher consanguineous with L. Wittgenstein. It’s certainly a killer argument.